Professional Documents
Culture Documents
F. E. M. Lilley*
Manuscript received by the Editor October 23, 1991; revised manuscript received March 3, 1993.
*Research School of Earth Sciences, Australian National University, GPO Box 4, Canberra A.C.T. 2601, Australia.
© 1993 Society of Exploration Geophysicists. All rights reserved.
1498
Magnetotelluric Analysis 1499
(1)
with
and
Then one-dimensionality of electrical conductivity structure, FIG. 1. Graphic representation of the values taken by the real
for which elements of a magnetotelluric impedance tensor upon rota-
tion of the measuring axes through angle 8’; (a) for induction
a = b, in 2-D structure, and (b) for induction in general 3-D struc-
is given by ture. Similar diagrams can be drawn for the quadrature
elements. Note that the notations and of Lilley (1976)
here become and and that tan is a skew value
for the real part of the tensor by itself. For 1-D structure, the
and extreme anisotropy, for which circle in (a) reduces to its central point.
1500 Lilley
For a 1-D case, tion states, not necessarily orthogonal to each other, for
which the electric and magnetic fields have the same polar-
b = a, ization at perpendicular orientations (see also the treatment
and of this problem by Yee and Paulson, 1987). When real and
quadrature parts of an impedance tensor are considered
separately, the ellipses sought by Eggers reduce to linear
polarizations, and their azimuths can be seen by inspection
and the circle reduces to its central point. of the appropriate Mohr circles.
The angle (clockwise) necessary to rotate the measuring The tensor quoted by Eggers (1982) is thus shown in
axes for alignment with the principal axes (to give the two Figure 2, with real and quadrature circles. What may now be
points where the circle intercepts the is also part sought, for each circle, are two directions for which the
of Figure la, entering as a rotation of the radial arm by angle electric and magnetic fields are at perpendicular orientations.
28 (anticlockwise) from its initial observed position at Taking the real circle first, these directions are clearly
= 0. those corresponding to the values of for which the circle
cuts the for it is at these points that is
A similar circle will also exist for the quadrature parts of a zero, and so an applied magnetic field will produce an
and b , with center 0) and radius The real and electric field E’y orthogonal to it. The actual values of are
quadrature circles will be distinct from each other; however, given in the figure. Two other values of also given in the
for a true 2-D case they must show the same angle 28 for figure, are similarly found for the quadrature circle.
rotation to 2-D alignment.
CONNECTION WITH THE LATORRACA ET AL. ANALYSIS
Circles for data from a 3-D structure
For a general case of a tensor indicating 3-D electrical LaTorraca et al. (1986) considered rotation of a complex
conductivity structure, the circle is offset from the Z'xy- impedance tensor in a different way, that is to obtain two
axis, as shown in Figure lb. The offset can be expressed by mutually orthogonal magnetic field directions for which the
an angle as shown. The value of tan is of a form similar associated electric fields are orthogonal to each other (but
to the “skew” of Swift (1967), and the offset angle shows not necessarily orthogonal to the magnetic fields).
the skew of the real part of the tensor graphically. Considering real and quadrature parts separately, and the
real part first, the LaTorraca directions are shown most
CONNECTION WITH THE EGGERS ANALYSIS directly on a plot of against , as in Figure 3.
Then, two points of fixed ratio [the LaTorraca
1982) considered rotation of the impedance tensor et al. (1986) condition for the direction of fields due to an
and (using an eigenstate formulation) sought two applied are given by the intersections with the circle of
anisotropy. As the = 0 radial arms of the ‘circles are not- parallel, the real and quadrature principal axes will not be in
alignment. The intercepts of the circles with the Z'xy,- and Z'xy,q-axes, at and as marked, indicate the appropriate
rotations (values of for the Eggers polarization states, taking real and quadrature parts separately.
Magnetotelluric Analysis 1501
a line from the origin to the circle’s center, continued to cut case is always positive, always negative, and
the circle a second time. Such constructions are shown in and Z’yy, range both positive and negative. The circle
Figure 3 for both the real and quadrature parts of the Eggers has radius and distances , and are as
tensor example, and the values of the rotation angle thus the
found are given. These values of can be compared with the With reference to Figure 4, by the theorem of Pythagorus
values determined in Figure 2 for the Eggers directions; the it is evident that
discrepancies are a consequence of the 3-D nature of the
data. = +
Clearly, for a 2-D tensor, the Eggers and et al. and
directions agree and become the directions of the 2-D
principal axes. = +
= = +
which is commonly called the determinantal impedance, and and
= RQ = + +
which is sometimes known as the “average” or “trace” it follows that
impedance. Applying these expressions to the real and quadra-
= .
ture parts of a tensor separately (as distinct from taking the real
and quadrature parts of and produces quantities that Thus, the distance on Figure 4, an invariant for the circle,
are also invariant, and which may be identified on the Mohr can be seen to represent the “first effective” or
circle depiction of an impedance tensor, as will now be shown. tal impedance, when only real parts of the tensor elements
The real part of the Eggers tensor is shown again as an are considered. Similarly the distance on Figure 4 repre-
example in Figure 4, where now axes are included for all of sents the “second effective” impedance, taking real parts of
Z’xy,, and Z’yy,, following the symmetry the tensor elements only. (In this case, because of the
relations indicated in Lilley (1976). Thus, for any point on simplicity of the expression, is also the real part of the full
the circle (corresponding to some particular axis rotation), complex second effective’ impedance.) As can be seen on
values for all the tensor elements can be scaled off. In this the figure, is greater or less than depending on whether
FIG. 3. The Eggers tensor plotted as circles of Z’xx versus Z’yx, with constructions shown to give the angles of rotation for
the magnetic field directions of LaTorraca et al. (1986). These angles are and for the real part of the tensor (left), and
and for the quadrature part (right).
1502 Lilley
the radius of the circle is greater or less than d4 (which is in Analogous to skew angle having a tangent of in
some respects a balance of anisotropy against skew). Figure 4, angle ROQ is noted to be an “anisotropy angle,”
A diagram similar to Figure 4 may also be drawn for the , equal to arctan That is,
quadrature elements of a tensor, with its own invariant
distances that correspond to and d2 on Figure 4. =
A similar index and angle exist for the quadrature part
A new invariant, the central impedance of a tensor.
Inspection of Figure 4 suggests that the distance from the For the Eggers (1982) tensor of Figure 2, the values of the
origin to the center of the circle, line OQ, is another anisotropy indices and angles are:
significant invariant. In terms of the Mohr circle representa- = 0.76, = 37”
tion, the center of the circle is the most central invariant
point of the tensor figure. Thus the length d3 of the line OQ, = 0.48, =
given by with skew angles being
+ + = and = 3”.
is a basic invariant impedance, and is termed here the Note that for this same complex tensor Eggers (1982) ob-
“central impedance.” There is also a central impedance tained a skew value of 12 degrees and LaTorraca et al. (1986)
value for the quadrature part of the tensor, with subscripts q a skew value of 11 degrees. These values fall within the
replacing subscripts r in the expression just given. range of and determined above.
An anisotropy angle FREQUENCY DEPENDENCE
Further, in Figure 4 the circle radius RQ is a natural An important aspect of observed magnetotelluric data is
measure of the tensor anisotropy, perhaps best normalized their frequency dependence. The Mohr circle method de-
by the distance d2. Thus an anisotropy index pr for the scribed so far is for data at a particular frequency, (or
tensor is evident as averaged over some frequency band). Data over a full range
= of frequencies may be displayed by color graphics using a
color spectrum to cover the frequency range. This method is
1 displayed here in principle, in Figure 5.
2 +
Figure 5 shows as an example, magnetotelluric data over a
+ range of four frequencies. The data are from a seafloor site
Index thus ranges in value from zero for a 1-D case (TP3) in the Tasman Sea recorded by the instruments of
(consider the circle in Figure la reduced to its central point) Filloux (1987) and reduced by Ferguson (1988). The experi-
to infinity (the consequence, as explained in the introduction ment is described in Lilley et al. (1989). Just four frequencies
above) of are shown for clarity; the values of the observed tensor
elements have been multiplied by a factor of each case
(where T is the period in hours) to normalize them against
FIG. 4. The real part of the Eggers tensor presented as a circle, with axes for all four tensor elements included.
The figure shows the construction to give d2, the magnitude of the determinantal impedance (considering only
real parts). The second effective (or “trace”) impedance is of magnitude dl . The “central impedance,” now
proposed as a basic invariant is d3 , the line joining the origin to the center of the circle. Angle is the skew
angle, and angle the “anisotropy angle.”
Magnetotelluric Analysis 1503
the characteristic that, above a uniform half-space, such 3) element t is the twist parameter which summarizes the
tensor elements are proportional to T twist effect on the regional tensor by multiplying it by
The circles show clearly that the observed data are 2-D in
character and highly anisotropic. (These points are impor- (1 + .
tant for marine data, as discussed by Ferguson et al. 1985, 1
1990, and Heinson, 1991.) Phase information is also evident Given observed values of , and Zyy , in the
in the relative sizes of the real and quadrature circles Groom-Bailey scheme the real and quadrature equations (2)
changing with frequency. are solved for , e , and a; i.e., eight equations for seven
unknowns (as and are complex).
APPLICATION TO GROOM-BAILEY DECOMPOSITION In this paper now note that equations (2) may be expressed
Groom and Bailey (1989) consider the case where a tensor = A + tan C) (3a)
that is regionally 2-D as in equation (1) is distorted locally by
shear, twist, and anisotropy. Introducing notation = tan A tan cos 28
= + zyy (tan A tan sin (3b)
= + = tan A tan C) (3c)
= a3 = A + tan cos
= zyy (1 tan A tan sin (3d)
they show that these processes affect a measured magneto- where A denotes the Groom-Bailey “twist” angle, so that
telluric impedance tensor
tan A = t
and the further notation has been introduced of
1
such that tan = =
= to + and
(2a)
tan C = =
eta) cos + ea) sin 28 (2b)
In Figure 6 below, notations are also used of
= + et8 (2c)
tan =
= + eo) cos 28 (6 eta) sin (2d)
and
where
tan = e.
1) the measurement axes are rotated angle from the Here it is generally understood that analysis proceeds sepa-
regional 2-D axes rately for the real and quadrature parts, with appropriate
2) parameter e is the shear parameter which affects the subscripts understood: thus, for example, = and
regional 2-D tensor as multiplication by a factor of tan = (note is not the real part of where and
(1 + and are complex).
FIG. 5. Magnetotelluric impedance data from a seafloor site in the Tasman Sea from Ferguson (1988) taken as an example for
plotting over a frequency range. The periods are 0.19. 0.48, 1.44, and 4.44 hr, and increasing period corresp ds to increasing
length of dashed line. Before plotting, the impedance values have been normalized by multiplication by T.
1504 Lilley
FIG. 6. Set of diagrams, of the form of Figure lb, for both general and particular cases of a Groom-Bailey magnetotelluric
impedance tensor. a) Shear only; b) Twist only; c) Both shear and twist. Except where stated otherwise, the examples have
been drawn taking 6 = 0.23, = 0.43, e = 0.58, and = 0.18. Each axis in the figure covers the range -0.1 units to +0.4 units.
In this figure, r denotes circle radius.
Magnetotelluric Analysis 1505
Some algebra then shows that 3) Both shear and twist; Figure 6, column (c).-The most
general case is presented in Figure 6c row (1). The radius of
tan (A + C) = the circle is enlarged from by a factor f where
so that, when defining skew angles for the real and quadra-
(1 + + = A sec B.
ture parts of a tensor separately, these will be given by (A +
C,) and (A + The center shifts to coordinates
It also follows from the above equations that
+
tan (A + B + 28) =
and so is offset from the horizontal axis by angle (A + C).
Depicting the decomposition graphically The 1-D case in Figure 6c row (2) is of regular 3-D form.
The high-anisotropy case, Figure 6c row (3), again goes
The relationships of equations (2) and (3) can be directly
through the origin, as does the strong or 2-D distortion case,
expressed by the diagrams in Figure 1. A suite of such
diagrams is presented in Figure 6, for both general and partic- Figure 6c row (5).
ular cases. It is instructive if shear and twist are introduced Comments on the principal impedances
separately, and then combined, and in Figure 6 this procedure
is adopted in columns (a), (b), and (c), respectively. With Groom and Bailey (1989), it is of interest to pursue a
There are five rows in Figure 6: row (1) for general cases, traditional 2-D search for recovery of the principal imped-
and rows (2-5) for particular cases considered by Groom and ances, carried out by minimizing
Bailey (1989) which may have distinctive characteristics.
+
Some of the main features of Figure 6 are now discussed.
as a function of coordinate rotation angle Expanded, this
1) Shear but no twist (e 0, 0); Figure 6, column procedure is the minimization of
(a).-The general case for shear only is shown in Figure 6a
row (1). Note that, relative to the undistorted 2-D tensor in + + +
Figure la, the center of the circle has been offset vertically, by Considering only the first and third terms, and with reference
an amount defined by the angle C shown. The radius of the to Figure 6c row (l), is a minimum where the circle
circle is enlarged from its undistorted value of by factor crosses the However, for a given position on the
f, where circle of , Z’yy , has a magnitude given by the value of
(1 + = that is diametrically opposite (see Figure 4). Hence
the minimum value of + will be given by
For 1-D structure, the circle is as shown in Figure 6a row the points of maximum and minimum , values. These
(2): of radius (l/2) and with the center on the Z’xy-axis. particular values, from inspection of the figure, are
Angle C is zero, and the diagram has the characteristics of a
2-D case. + A B,
For the contrasting special case of high anisotropy, the circle
and
takes the form shown in Figure 6a row (3). This circle is
distinctive in going through the origin, (a result of the high A B.
anisotropy approximation which in effect sets to zero; in
practice, such a circle would be close to the origin, but not Similar results hold for the minimization of +
through it).
For weak distortion, Figure 6a row (4), the values of The values of rotation angle at which the maximum and
and can be read directly off the graph, with little approx- minimum values of occur can be found by seeking
imation; and for strong distortion, characterized in Figure 6a maximum and minimum values of the function of Lilley
row (5) with 1, values of a and S are again given (1976), which in the notation of the present paper means
graphically. seeking maximum and minimum values of a function
= sin 26’.
2) Twist but no shear 0, e = 0); Figure 6, column
(b).-The general case for twist only is shown in Figure 6b Some algebra produces the result, using equations (3b) and
row (1). Again, relative to the undistorted tensor of Figure la (3d), that
the center of the circle has been offset vertically, by twist
= A B cos (A + B + 28 28’)
angle A. The radius of the circle has been enlarged by factor
sec A. so that the maximum and minimum values of can be
For a twist with a 1-D case, Figure 6b row (2), there is the seen to occur when
unusual result of an offset single point, so that rotation of
....
axes produces no effect. As is general for column 6b, the
offset is by the twist angle A. i.e., when
High anisotropy, Figure 6b row (3), brings the circle
through the origin; and weak distortion, Figure 6b row (4),
produces a circle of simple description, from which values of Thus when real and quadrature parts are considered sepa-
S, and can be read. rately, a general version of Groom and Bailey’s (1989) result
1506 Lilley
(their equation 41a) is obtained: that the directions found by ACKNOWLEDGMENTS
minimizing and maximizing , will differ from the prin-
cipal strike directions by (1/2)(A + B). Note that generally The author acknowledges beneficial discussions on mag-
(l/2) (A + B) is not half the skew angle, which is (112) netotelluric matters with Jean H. Filloux, Ian J. Ferguson,
(A + C). In the high-anisotropy case considered by Groom and Graham S. Heinson. Richard C. Bailey made helpful
and Bailey, for then B = C. comments on an earlier version of the manuscript, and the
Also, of course, the maximum and minimum values found reports of the referees have been of great value. Win Means
for , , as given above, are not the regional 2-D values of is thanked for stimulating discussion of Mohr circles.
+ and [( there appears to be no
direct graphical method for identifying these latter (though REFERENCES
they can be read off the graph in some particular cases such
Bahr, K., 1988, Interpretation of the magnetotelluric impedance
as those of weak distortion, given in Figure 6). tensor: Regional induction and local telluric distortion: J. Geo-
phys., 62, 119-127.
Discussion - 1991, Geological noise in magnetotelluric data: A classifica-
tion of distortion types: Phys. Earth Planet. Int., 66, 24-38.
The diagrams in Figure 6 make some particular points as Berdichevsky, M. N., and Dmitriev, V. I., 1976, Basic principles of
interpretation of magnetotelluric curves, in Adam., Ed., Geoelec-
follows. tric and geothermal studies: Akademiai Kaido Budapest, KAPG
Geophys. Mono., 165-221.
Counil, J. L., LeMouel, J. L., and Menvielle, M., 1986, Associate
1) A locally-sheared but regional 1-D case, as in Figure 6a and conjugate directions concepts in magnetotellurics: Ann. Geo-
row (2), is indistinguishable from a simple regional 2-D phys., B4, 115-130.
case, as in Figure la. Eggers, D. E., 1982, An eigenstate formulation of the magnetotel-
luric impedance tensor: Geophysics, 47, 12061214.
2) A locally-twisted but regional 1-D case, Figure 6b row Ferguson, I. J., 1988, The Tasman project of seafloor magnetotel-
(2), produces a distinctive simple point plot. luric exploration: Ph.D. thesis, Australian National University,
3) High regional 2-D anisotropy and strong local 2-D Ferguson, I: J., Filloux, J. H., Lilley, F. E. M., Bindoff, N. L., and
distortion both cause the circle to go near the origin, so Mulheam, P. J., 1985, A seafloor magnetotelluric sounding in the
there is one orientation of axes that gives low values for Tasman Sea: Geophys. Res. Lett., 12, 545-548.
Ferguson, I. J., Lilley, F. E. M., and Filloux, J. H., 1990, Geomag-
both Z'xx and Z'xy. netic induction in the Tasman Sea and electrical conductivity
4) In view of the discussion concerning the invariants in structure beneath the Tasman seaflloor: Geophys. J. Int., 102,
Figure 4, Figure 6 shows that both high anisotropy and 299-312.
Filloux, J. H., 1987, Instrumentation and experimental methods for
strong or 2-D distortion will have the effect of reducing oceanic studies, in Jacobs, J. A., Ed., Geomagnetism: Academic
the determinantal impedance (taking real parts of the Press, vol. 1, 143-247.
tensor elements only) to near zero, while the second Groom, R. W., and Bailey, R. C., 1989, Decomposition of magne-
totelluric impedance tensors in the presence of local three-
effective (or “trace”) impedance is changed much less dimensional galvanic distortion: J. Geophys. Res., 94, 1913-1925.
from its ideal value of (1/2)a. This point, for real and - 1991, Analytic investigations of the effects of near-surface
three-dimensional galvanic scatterers on MT tensor decomposi-
quadrature parts taken separately, may be relevant to tions: Geophysics, 56, 496-518.
the discussion in Groom and Bailey (1989) (and else- Heinson, G. S., 1991, Interpretation of seafloor magnetotelluric data
where) on the relative effects of distortion on the using thin-sheet modeling: Ph.D. thesis, Australian National
University, Canberra.
determinantal and trace impedances. Ingham, M. R., 1988, The use of invariant impedances in magneto-
telluric interpretation: Geophysical, J., 92, 165-169.
Generally, diagrams for the Groom-Bailey case of shear Jaeger, J. C., 1969, Elasticity, fracture and flow (3rd ed.): Methuen.
Jaeger, J. C., and Cook, N. G. W., 1976, Fundamentals of rock
and twist have the form of the 3-D case of Figure lb, with the mechanics (2nd Ed.): John Wiley & Sons, Inc.
implication that a typical 3-D tensor may be subjected to LaTorraca, G. A., Madden, T. R., and Korringa, J., 1986, An
Groom-Bailey analysis. The criterion for such analysis to be analysis of the magnetotelluric impedance for three-dimensional
conductivity structures: Geophysics, 51, 1819-1829.
valid is that both the real and quadrature parts of a tensor Lilley, F. E. M., 1976, Diagrams for magnetotelluric data: Geophys-
should yield the same set of e, t, and values. ics, 41, 766-770.
Lilley, F. E. M., Filloux, J. H., Ferguson, I. J., Bindoff, N. L., and
Mulheam, P. J., 1989, The Tasman project of seafloor magneto-
CONCLUSIONS telluric exploration: Experiment and observations: Phys. Earth
Planet, Int., 53, 405-421.
In their recent review, Wannamaker and Hohmann (1991) Nye, J. F., 1957, Physical properties of crystals: Oxford Press.
refer to a current revolution in electromagnetic induction Park, S. K., 1985, Distortions of magnetotelluric sounding curves by
three-dimensional structures: Geophysics, 50, 785-797.
studies. The magnetotelluric method has been central in this Park, S. K., and Livelybrooks, D. W., 1989, Quantitative interpre-
development, with data of increased precision observed over tation of rotationally invariant parameters in magnetotellurics:
Geophysics, 54, 1483-1490.
wide ranges of frequency. With modern computer software Spitz, S., 1985, The magnetotelluric impedance tensor properties
available for data reduction, there is still a need (indeed with respect to rotations: Geophysics, 50, 1610-1617.
perhaps an increased need) for a display of diagnostic Swift, C. M., Jr., 1967, A magnetotelluric investigation of an
electrical conductivity anomaly in the southwestern United
characteristics that help the judgements of an interpreter. States: Ph.D. thesis, Mass. Inst. of Tech.
The Mohr circle method may be used to display magneto- Wannamaker, P. E., and Hohmann, G. W., 1991, Electromagnetic
telluric data succinctly, in the same way that it has proved induction studies: Rev. Geophys. Supplement (U.S. Natl. Rep. to
IUGG 1987-1990), 405-415.
useful for other tensor information. A visual impression is Yee, E., and Paulson, K. V., 1987, The canonical decomposition
given of two-dimensionality, three-dimensionality, anisot- and its relationship to other forms of magnetotelluric impedance
tensor analysis: J. Geophys., 61, 173-189.
ropy, and skew, and insight is gained into the various Zhang, P., Roberts, R. G., and Pedersen, L. B., 1987, Magnetotel-
magnetotelluric invariants and particular rotation angles. luric strike rules: Geophysics, 51, 267-278.